This is the first of a three part series by Matthew Heusser, software delivery consultant and writer.
Every now and again and opportunity comes along that you just can’t refuse. Mine was to teach the one-day version of my class, lean software testing, in Madrid, Spain, then again the following week in Estonia. Instead of coming back to the United States, I’ll be staying in Europe, with a few days in Scotland and a TestRetreat in the Netherlands.
And a lot of time on airplanes.
The folks at Sauce Labs thought I might like to take notes and type a little on the plane, to share my stories with you.
The first major hit in Madrid is the culture shock; this was my first conference where English was not the primary language. The sessions were split between English and Spanish, with translators in a booth making sure all talks were available in all languages.
The Testing Divide
Right now, in testing, I am interested in two major categories: The day to day work of testing new features and also the work of release-testing after code complete. I call this release testing a ‘cadence’, and, across the board, I see companies trying to compress the cadence.
My second major surprise in Madrid is how wide the gap is —and I believe it is getting wider —between legacy teams that have not modernized and teams starting from scratch today. One tester reported a four-month cycle for testing. Another team, relying heavily on Cucumber and Selenium, were able to release every day.
Of course, things weren’t that simple. The Lithuanian team used a variety of techniques I can talk about in another post to reduce risk, something like devOps, which I can talk about in another post. The point here is the divide between the two worlds.
Large cadences slow down delivery. They slow it down a lot; think of the difference between machine farming in the early 20th century and the plow and horse of the 19th.
In farming, the Amish managed to survive by maintaining a simple life, with no cars, car insurance, gasoline, or even electricity to pay for. In software, organizations that have a decades-long head start: banks, insurance companies, and pension funds, may be able to survive without modernization.
I just can’t imagine it will be much fun.
Batches, Queues and Throughput
Like many other conferences, the first day of ExpoQA is tutorial day, and I taught the one-day version of my course on lean software testing. I expected to learn a little about course delivery, but not a lot —so the learning hit me like a ton a bricks.
The course covers the seven wastes of ‘lean’, along with methods to improve the flow of the team - for example, decreasing the size of the measured work, or ‘batch size’. Agile software development gets us this for free, moving from ‘projects' to sprints, and within sprints, stories.
In the early afternoon we use dice and cards to simulate a software team that has equally weighted capacity between analysis, dev, test and operations —but high variability in work size. This slows down delivery. The fix is to reduce the variation, but it is not part of the project, so what the teams tend to do is to build up queues of work, so any role never runs out of work.
What this actually does is run up the work in progress inventory - the amount of work sitting around, waiting to be done. In the simulation I don’t penalize teams for this, but on real software projects, ‘holding’ work created multitasking, handoffs, and restarts, all of which slow down delivery.
My lesson: Things that are invisible look free —and my simulation is far from perfect.
After my tutorial it is time for a conference day - kicked off by Dr. Stuart Reid, presenting on the new ISO standard for software testing. Looking at the schedule, I see a familiar name; Mais Tawfik, who I met at WOPR20.Mais is an independent performance consultant; today she is presenting on “shades of performance testing.”
Performance Test Types
Starting with the idea that performance testing has three main measurements: Speed, Scalability, and Stability, Mais explains that there are different types of performance tests, from front-end performance (javascript, waterfalls of HTTP requests, page loading and rendering) to back-end (database, webserver), and also synthetic monitoring - creating known-value transactions continuously in production to see how long they take. She also talks about application usage patterns - how testing is tailored to the type of user, and how each new release might have new and different risks based on changes introduced. That means you might tailor the performance testing to the release.
At the end of her talk, Mais lists several scenarios and asks the audience what type of performance test would blend efficiency and effectiveness. For example, if a release is entirely database changes, and time is constrained, you might not execute your full performance testing suite/scripts, but instead focus on rerunning and timing the database performance. If the focus on changes is the front end, you might focus on how long it takes the user interface to load and display.
When Mais asks if people in the organization do performance testing or manage it, only a handful of people raise their hands. When she asks who has heard of FireBug, even less raise their hand.
Which makes me wonder if the audience is only doing functional testing. If they are, who does the performance testing? And do they not automate, or do they all use Internet Explorer?
The talk is translated; it is possible that more people know these tools, it was just that the translator was ‘behind’ and they did not know to raise their hands in time.
Here’s hoping!
Time For A Panel
At the end of the day I am invited to sit on a panel to discuss the present (and future) of testing, with Dr. Reid, Dorothy Graham, Derk-Jan De Grood, Celestina Bianco and Delores Ornia. The questions include, in no particular order:
Will testers have to learn to code?
How do we convince management of the important of QA and get included in projects?
What is the future of testing? Will testers be out of a job?
What can we do about the dearth of testing education in the world today?
For the problem with the lack of education, Dorothy Graham points to Dr. Reid and his standards effort as a possible input for university education.
When it is my turn, I bring up ISTQB The International Software Testing Qualifications Board. - if ISTQB is so successful (“300,000 testers can’t be wrong?”) then why is the last question relevant? Stefaan Luckermans, the moderator, replied that with 2.9 Million testers in the world, the certification had only reached 10%, and that’s fair, I suppose. Still, I’m not excited about the quality of testers that ISTQB turns out.
The thing I did not get to say, because of time, that I want to do is point out that ISTQB is, after all, just a response to a market demand for a 2-3 day training certification. What can a trainer really do in 2-3 days? At most, maybe, teach a single technical tool, turn the lightbulb of thinking on, or define a few terms. ISTQB defines a few terms, and it takes a few days.
The pursuit of excellent testing?
That’s the game of a lifetime.
By Matthew Heusser - matt.heusser@gmail.comfor Sauce Labs
Stay tuned next week for part two of this mini series! You can follow Matt on Twitter at @mheusser.
Have an idea for a blog post, webinar, or more? We want to hear from you! Submit topic ideas (or questions!) here.
